Stark Effect Spectroscopy of Tryptophan (AV)- 2 ... - Stanford University
1 583
Biophysical Journal Volume 68 April 1995 1583-1591
Stark Effect Spectroscopy of Tryptophan Daniel W. Pierce and Steven G. Boxer Department of Chemistry, Stanford University, Stanford, California 94305 USA
ABSTRACT The change in permanent dipole moment (IA1p I) for the transition from the 1La state to the ground state of tryptophan is the key photophysical parameter for the interpretation of tryptophan fluorescence spectra in terms of static and dynamic dielectric properties of the surrounding medium. We report measurement of this parameter by means of electric field effect (Stark) spectroscopy for N-acetyl-L-tryptophanamide (NATA) in two solvents, the single tryptophan containing peptide melittin, and 5-methoxytryptophan. The values ranged from 5.9 to 6.2 ± 0.4 Debye/f for NATA and melittin, where f represents the local field correction. The 1Lb AlP was much smaller. Application of Stark spectroscopy to these chromophores required decomposition of the near-UV absorption into the 1La and 'Lb bands by measurement of the fluorescence excitation anisotropy spectrum and represents an extension of the method to systems where band overlap would normally preclude quantitative analysis of the Stark spectrum. The results obtained for 5-methoxytryptophan point out limitations of this method of spectral decomposition. The relevance of these results to the interpretation of steady-state and time-resolved spectroscopy of tryptophan is discussed.
INTRODUCTION Measurement of steady-state and time-resolved fluorescence of tryptophan is widely used for characterization of peptides and proteins, and results are routinely interpreted in terms of the static polarity and heterogeneity of the local environment as well as the dynamic properties of the environment and the emitting residue (Beecham and Brand, 1985). The key photophysical parameter for the interpretation of fluorescence in terms of local polarity is the change in permanent dipole moment ( A'i ) on absorption. The classical method for measuring this quantity is to measure the fluorescence spectrum in a series of solvents of known polarity. The difference dipole is given by the Ooshika-Lippert-Mataga relation
(Lippert, 1957):
(AV)- 2("ie pg)
(r
I
n
2
1
(1)
where (A v) (A vA) - (AVF), and (AvA) and (AvF) are the shifts of the 0-0 absorption and fluorescence transition energies, respectively, due to electrostatic interactions with the medium. These shifts are measured relative to the values obtained in a nonpolar solvent. The quantities Me and Mg are the excited- and ground-state permanent dipole moments, and A'i = Me- Mg. E0 is the static dielectric constant, n is the refractive index, a is the cavity radius of the solvent cavity (typically assumed to be spherical) required to accommodate the chromophore, and h is Planck's constant. This equation remains extremely useful but suffers from the same shortcomings as all relations based on a continuum dielectric model of a solvent: all perturbations due to specific inter-
Received for publication 28 July 1994 and in final form J0 January 1995. Address reprint requests to Dr. Steven G. Boxer, Department of Chemistry, Stanford University, Stanford, CA 94305-5080. Tel.: 415-723-4482; Fax: 415-723-4817; E-mail: [email protected]. C) 1995 by the Biophysical Society
0006-3495/95/04/1583/09 $2.00
actions between solute and solvent (i.e., ground- or excitedstate complex formation) and to the finite size of the solvent molecules are ignored. Such effects will adversely impact the accuracy of the determination of AM I . In addition, Me is assumed to be the same in the Franck-Condon excited state and the relaxed emitting state, and obtaining reliable absorption and fluorescence spectra of dipolar or charged molecules in a truly nonpolar reference solvent can be challenging because of limited solubility and aggregation. A better method, in principle, is to measure the effect of an externally applied electric field on the electronic transition (the Stark effect; Liptay, 1976). The electric field changes the wavelengthdependent molar absorption coefficient in a manner dependent on A' I, the change in the average polarizability (TrAa), and the internal angle ; between the transition moment 'm and AM,. Here we report the results of Stark effect and fluorescence excitation anisotropy studies at low temperature (77-85 K) on N-acetyl-L-tryptophanamide (NATA) in ethanol (EtOH) or 50% v/v glycerol/water glass, melittin in 50% glycerol/water, and 5-methoxytryptophan in EtOH. These molecules were selected to examine the sensitivity of the spectra to different environments and to chemical substitution. The near-UV absorption of tryptophan is believed to consist of two nearly totally overlapping T-u7T* transitions, labeled 'La and 1Lb in the Platt notation (Valeur and Weber, 1977). The formalism for the analysis of Stark spectra developed by Liptay (1976) rests on the assumption that the absorption lineshape is that of a single, isolated electronic transition with wavelength-independent electrooptic properties. This assumption is necessary because the Stark spectrum is expressed in terms of sums of derivatives of the absorption lineshape. However, there is no fundamental obstacle to analysis of the Stark spectrum of overlapping transitions as long as there is some means of decomposing the total absorption into its constituent parts. In the case of
1584
Biophysical Journal
tryptophan, the observation that the fluorescence originates solely from 'La and the near orthogonality of the 'La and 'L4 transition moment directions provides a means for accomplishing such decomposition, performed originally by Valeur and Weber (1977). The emission anisotropy at differing excitation wavelengths is measured and interpreted in terms of the fractional absorption accounted for by each transition moment at each excitation wavelength. 5-Methoxytryptophan was selected because substitution of electronwithdrawing moieties at position 5 is thought to selectively lower the energy of the '4 transition and switch the order of 0-0 energies in both absorption and fluorescence without perturbation of 'La, and previous work has indicated that this is the only methoxy derivative that retains orthogonality of the transition moments (Eftink et al., 1990).
MATERIALS AND METHODS NATA was a gift of David J. Lockhart. Melittin and 5-methoxytryptophan were purchased from Sigma Chemical Co. (St. Louis, MO) and used without further purification. The glycerol was Mallinkrodt AR grade and the EtOH was Aldrich spectrophotometric grade.
Stark and absorption spectra The Stark apparatus used in these experiments has been described elsewhere (Boxer, 1993). Light from a 450-W Xe arc lamp (Spex) was dispersed through a Spex Minimate ¼/4-m monochromator (0.9-nm resolution), collimated, passed through an air-spaced Glan-Taylor polarizer and a strain-free quartz optical dewar containing the sample immersed in liquid nitrogen, and detected by a photomultiplier tube on which the resistor network had been altered so as to bypass the last few dynodes and decrease the noise. The photomultiplier signal was converted to a voltage and amplified using a home-built converter/amplifier and fed into a lock-in amplifier (Stanford Research Systems SR530). Samples consisted of 15 ,ul of solution between two quartz slides held apart by a 25-,um Kapton spacer. The quartz slides were rendered conductive by means of an evaporated layer of nickel -7.5 nm thick, corresponding to a transmittance of about 50% per slide in the near UV and visible and a resistance of not more than 1000 Q. 400-Hz sinusoidal AC electric fields with peak-to-peak voltages from 2 to 4.5 kV were produced by a custom-built power supply and applied to the sample. The raw data are the intensity of the signal at the photomultiplier, I(v), and the amplitude of its modulation at twice the field frequency, AI(v). The Stark spectrum is AI(v)/(I(v)-k), where k is a constant baseline voltage from the amplifier. Because the electric field modulation is small (1150 K. Although this energy will rapidly flow into a large number of low-frequency vibrational modes, perhaps enough local heating occurs so that small rotations of the molecule may occur. An analogous process could occur on emission, but by the time a vibrationally hot ground state is created the photon is gone and any rotation of the ground-state molecule will not be observed. In either case, the validity of the absorption decomposition given by Eq. 4 must be called into question if such decomposition is carried out over multiple vibronic maxima of ei-
z Fm .
5-methoxytryptophan
ca
0
0.05 V
0
0.0
0.00
0-
32000 33000 34000 35000
-0.1
21000 24000 27000 30000 ENERGY
cm
l
ENERGY cm-1
FIGURE 6 Absorption (A) and fluorescence excitation anisotropy (V) spectra for 5-methoxytryptophan at 77 K in EtOH. The absorption data is identical to that in Fig. 5 except that a smaller range of data is shown. Solid lines do not show fits in this figure.
FIGURE 7 5-Methoxytryptophan singlet (A) and triplet (A) emission and emission anisotropy spectra. The fluorescence and phosphorescence spectra have been independently scaled to a peak intensity of 1. The spectra have been corrected for the wavelength and polarization-dependent efficiency of the detection apparatus.
Pierce and Boxer
1 589
Stark Effect Spectroscopy of Tryptophan
ther transition. For 5-methoxytryptophan, there does not appear to be a clear-cut means of identifying the onset of 'La absorption from r,(v), so we analyze the Stark spectrum of the first vibronic band in the normal manner for an isolated electronic transition. For NATA and melittin, the excitation anisotropy spectra observed are more plausibly attributed to overlapping electronic transitions. To avoid running afoul of assumption 1, we restrict our analysis of the Stark spectra of these molecules to the region containing the first vibronic band of each transition (32,000-35,000 cm-').
DISCUSSION Comparison with solvent-dependence data Mataga and co-workers (1964) estimated Ai for the 'La state of indole to be 5 Debye (D) on the basis of the solvent dependence of the emission spectrum, and Gladchenko and Pikulik (1967) estimated jie to be 5.6 D based on the temperature dependence of absorption and emission spectra. These results were called into question by Walker et al. (1967), who interpreted at least part of the solvent dependence to exciplex formation in protic solvents. Lami and Glasser (1985) undertook a more contemporary and detailed study that resolved several points of confusion in the literature. First, they demonstrated that the identity of the emitting state depends on solvent polarity for indole and several derivatives, which presents a clear problem in the interpretation of solvent-dependent emission data. Second, evidence for ground-state complex formation in all polar solvents tested leading to large shifts in the 'La energy was obtained. Third, solvent effects on absorption and emission were separately evaluated. However, ground-state complex formation precluded independent determination of A'i for 'La absorption and emission, so the assumption of no change in je on relaxation of the Franck-Condon excited state still had to be made. Under this assumption, for the 'La state of 3-methylindole was found to be 5.32 D, and that of the 1Lb state of 5-methoxyindole 2.69 D. The corresponding groundstate values were 2.10 and 2.0 D, leading to A'i values of 3.22 and 1.69 D, respectively. This value for the 'La Api is smaller than that obtained in this work. The difference is too large to be accounted for by any plausible estimate of the local field correction, and must therefore be ascribed to limitations in the accuracy of one or both of the methods being compared.
surements is qualitative at best. Condensed-phase molecular dynamics coupled with similar electronic structure calculations (Muino et al., 1992) indicate an even larger value for the 'La dipole. Chabalowski et al. (1993) performed ab initio calculations on indole including optimization of excited-state geometries and the effects of a reaction field. They found 'La and 1Lb lie values of 6.74 and 3.17, respectively, and a igI of 2.94 D after two iterations at the ground-state geometry and reaction field, although the convergence of the calculation after this number of iterations is unknown. However, the corresponding values for the fluorescence transition at the excited-state geometry and reaction field were 10.4, 4.72, and 5.41 D, respectively. Levy and co-workers have performed extensive vacuum and condensed phase ab initio electronic structure calculations with great attention to the effects of the solvent field on the properties of electronic states (Levy et al., 1991; Westbrook et al., 1992). The change in dipole moment on excitation to the 'La state predicted by their work is 3.1 D. This value represents an ensemble average over many configurations of solvent. The range of values of l for the configurations sampled was 2.4 to 9.0 D (K. Krogh-Jespersen, personal communication), and the excited-state dipole moment was found to increase rapidly to 14 D in simulated water at 300 K as the water reorganized to solvate the excited-state dipole moment, and the excitedstate dipole moment in turn changed in response to the reaction field due to the water. A common theme in all of these computations is the large change in the charge distribution of the 'La state that occurs during the excited-state lifetime in polar solvents. The magnitude of these changes illustrates the displaced nature of the 'La transition and the hazards present in assuming that photophysical properties are independent of time and environment (as is assumed in the classical solvent shift relations.)
Inhomogeneous effects and solvent dependence The Stark lineshape data in this work are analyzed under the assumption that the electrooptic properties of the molecule are not inhomogeneously distributed, or more precisely that the width of the distribution in a parameter is small compared with the magnitude of the parameter. The Stark signal is quadratic in Ali and Aa I, so the quantities determined are actually the square root of the ensemble average of the square of the quantity. For example, if Ali is characterized by a gaussian distribution with width ar and center value POiA the measured value is given by
Comparison with electronic structure calculations Callis (1991) performed extensive semiempirical molecular orbital calculations on indole and was able to reproduce many experimentally observed features. Calculated 'La l values ranged from 3.22 to 7.07 D but were generally about 5 D. A gas-phase value for Api of 5.7 D for 3-methylindole is also reported by Muinlo et al. (1992). However, the relevance of gas-phase calculations to condensed-phase mea-
2 -I AAo1)2/a2 1 '2 r(/ApiI= I-IA01)2/or2 (I -) f oI Ap' (IAAe-(lAA'
f
00
(12)
e
so
(IAp2 (15p1)
_
V&±IA 0-I2 IA"*
(13)
1 590
Biophysical Journal
where angle brackets denote an ensemble average. For instance, if A'I is 5 D and or is 2 D, the measured value would be 5.39 D. If there is a correlation between the value of an electrooptic parameter and the transition energy of a chromophore, the measured value will also depend on the region of the (inhomogeneously broadened) spectrum that is analyzed (Demchenko and Ladokhin, 1988). An interesting aspect of recent condensed-phase electronic structure calculations is the possibility of evaluating such correlations. Large variations in the 'La dipole moment were predicted in several of the studies described above (Muifio et al., 1992; K. Krogh-Jespersen, personal communication), and in one of them (Muifio et el., 1992) a correlation was predicted, with red-absorbing solvent configurations characterized by larger excited-state dipole moments. Because we focus on the red part (although not the extreme low-energy tail that is usually the subject of "red-edge effect" investigations) of the absorption band, this experiment may be sensitive to this bias (Demchenko and Ladokhin, 1988). However, because an entire vibronic maximum is analyzed for each transition, a width of the order of the inhomogeneous linewidth is included in the analysis, and any bias must be correspondingly small. An experimental approach to the correlation of electrooptic parameters with transition energy is provided by the combination of electric field effects with line-narrowing spectroscopy (Meixner et al., 1992; Altmann et al., 1992). The most directly relevant of the available studies is that of Vauthey et al. (1993). For highly polar dyes in highly polar polymer matrices, a substantial wavelength dependence to the amplitude and lineshape of the electric field effect was observed. The lineshape changes indicated that the angle between Ai and 'm changed as a function of wavelength. The data were interpreted as a rotation of Aji brought about by changing relative contributions of "intrinsic" and matrix-induced difference dipole moments to AjI, but we note that rotation ofthe transition moment due to vibronic coupling (see above) would provide an alternative mechanism. The values of Ai recovered for both transitions for NATA in EtOH, NATA in glycerol/water, and mellitin are identical within the errors despite the rather large changes in 0-0 transition energies and inhomogeneous linewidths seen in Figs. 2 and 4. The Stark spectra in Fig. 3 reflect the differences in absorption lineshapes but are otherwise identical. The solventindependence ofthe electrooptic properties is not consistent with significant inhomogeneous distortion of the recovered parameters. If the electrooptic properties are independent of solvent, they should be correspondingly insensitive to different local structures within the same solvent. It is interesting to note that the sample in the least polar solvent, NATA in EtOH, has the most redshifted 'La and 1, absorbance maxima, in contrast to the continuum electrostatic prediction. The 'La linewidths, however, do follow the expected trends with solvent polarity.
Resolution of the 1La and 'Lb transitions The decomposition of the 'La and 'Lb transitions in this work differs from that performed by Valeur and Weber (1977) in
Volume 68 April 1995
that the decomposed spectra were required to fit the Stark spectrum. Since this constraint was present, it was not necessary to assume a value for the limiting anisotropy of either transition, and these were left as free parameters in the fit. In the case of NATA and melittin, the assumption that the relatively flat region in the low-energy tail of the excitation anisotropy spectrum represents nearly pure 'La absorption was born out by the fitting. The limiting anisotropy determined for the 'La transition is less than the theoretical limiting value of 0.4 and indicates that the transition dipole moments on absorption and emission are not completely parallel. The values obtained (0.32-0.35) correspond to angles of 17-21°. Such rotations can derive from vibronic coupling (see above) or nuclear rearrangements during the excited-state lifetime. Since the original experiment of Valeur and Weber (1977), two other experimental approaches to the decomposition of indole absorption spectra have been pursued. Albinsson and Norden (1992) have measured linear dichroism spectra of indole and derivatives in stretched films, and Rehms and Callis (1987) have measured two-photon fluorescence excitation spectra. In addition, resolutions of a variety of derivatives have been performed by the fluorescence excitation anisotropy technique (Eftink et al., 1990). Of these, only the two-photon absorption measurements do not rely directly on wavelength-independent transition moment directions in order to measure relative 'La and 1Lb oscillator strengths. However, two-photon and one-photon lineshapes need not be identical, so there is at present no unambiguous means of decomposing the total absorption over the entire band. Our observations suggest that the available resolutions may not be even qualitatively useful except in the region of the 0-0 bands. As the analysis of Stark spectra depends rather sensitively on the absorption lineshape, it is likely that the largest nonstatistical errors in this experiment are due to violations of the assumptions necessary in the absorption decomposition.
CONCLUSION The most robust conclusion of this work is that A'i for the 'La transition of the indole chromophore in NATA and melittin is about 6.0 D/f. The long-wavelength absorption tail is due entirely to this transition, and the Stark spectrum in this region is dominated by a second-derivative contribution from it. It is therefore impossible to reproduce the experimental data by any other choice of parameters. The other parameters given in Table 1 are accurate within the assumptions necessary to decompose the absorption spectrum, but do not have as unique a relationship to the Stark spectrum. It can be stated that the 'Lb A'i cannot be much larger than the value given. The magnitude of the Stark effect is proportional to the inverse square of the linewidth of a gaussian transition. Given that the linewidth of the 'Lb 0-0 transition is much narrower than that of 'La, a larger A'i for 'Lb would have a
dramatic effect
the
'La
A'
we
on
the total Stark lineshape. The
obtain is
larger
value of than that calculated from
Pierce and Boxer
Stark Effect Spectroscopy of Tryptophan
solvent shifts or predicted by some electronic structure calculations. A value of the local field correctionf > 1.5 would be necessary to produce agreement with these calculated values. This value offis possible, but would be at the upper end of the likely range (Oh et al., 1991). These results affect the interpretation of tryptophan fluorescence properties in several ways. First, if it is accepted that the 'La A'i on absorption is larger than has been appreciated, quantitative conclusions as to the polarity of the environment based on absorption shifts must be rescaled. The situation with regard to fluorescence shifts is less clear, because the dipole moment may change significantly during the excited-state lifetime. Second, the evidence presented here for wavelength-dependent transition moment directions of one or both states must be taken into account in the interpretation of anisotropy-based measurements of conformational mobilities of tryptophan residues. For example, a rapid anisotropy decay component in the total emission could arise from vibrational relaxation, because different vibronic transitions have different transition moment directions. Such a component could easily be misinterpreted as deriving from motions or 'La-1Lb state dynamics. This work was supported in part by a grant from the National Institutes of Health D.W.P. was supported in part by a National Science Foundation predoctoral fellowship and by the Biophysics Training Grant at Stanford
University.
REFERENCES Albinsson, B., and B. Nord6n. 1992. Excited state properties of the indole chromophore: electronic transition moment directions from linear dichroism measurements: effect of methyl and methoxy substituents. J. Phys. Chem. 96:6204-6212. Albinsson, B., M. Kubista, B. Norden, and E. W. Thulstrup. 1989. Near-ultraviolet electronic transitions of the tryptophan chromophore: linear dichroism, fluorescence anisotropy, and magnetic circular dichroism spectra of some indole derivatives. J. Phys. Chem. 93:6646-6654. Albrecht, A. C. 1961. Polarizations and the assignments of transitions: the method of photoselection. J. Mol. Spectro. 6:84-108. Altmann, R. B., I. Renge, I. Kador, and D. Haarer. 1992. Dipole moment differences of nonpolar dyes in polymeric matrices: stark effect and photochemical hole burning. I. J. Chem. Phys. 97(8):5316-5322. Beechem, J. M., and L. Brand. 1985. Time-resolved fluorescence of proteins. Annu. Rev. Biochem. 54:43-71. Bottcher, C. J. F. 1973. Theory of Electric Polarization, 2nd ed, Vol. 1. Elsevier Scientific Publishing Company, Amsterdam. 77. Boxer, S. G. 1993. Photosynthetic reaction center spectroscopy and electron transfer dynamics in applied electric fields. In The Photosynthetic Reaction Center, Vol. II. J. Deisenhofer and J. R. Norris, editors. Academic Press, San Diego. 179-220. Boxer, S. G., A. Kuki, K. A. Wright, B. Katz, and N. H. Xuong. 1982. Oriented properties of the chlorophylls: Electronic absorption spectroscopy of orthorhombic pyrochlolophyllide a-apomyoglobin single crystals. Proc. Natl. Acad. Sci. USA. 79:1121-1125.
1591
Callis, P. R. 1991. Molecular orbital theory of the 'La and '4 states of indole. J. Chem. Phys. 95:4230-4240. Chabalowski, C. F., D. R. Garmer, J. 0. Jensen, and M. Krauss. 1993. Reaction field calculations of the spectral shifts of indole. J. Phys. Chem. 97:4608-4613. Demchenko, A. P., and A. S. Ladokhin. 1988. Red-edge-excitation fluorescence spectroscopy of indole and tryptophan. Eur. Biophys. J. 15: 369-379. Eftink, M. R., L. A. Selvidge, P. R. Callis, and A. A. Rehms. 1990. Photophysics of indole derivatives: experimental resolution of La and Lb transitions and comparison with theory. J. Phys. Chem. 94:3469-3479. Gladchenko, L. F., and L. G. Pikulik. 1967. Determination of dipole moments of indole and tryptophan in excited states. Zh. Prikl. Spektrosk. 6:355-60. Hill, N. E., W. E. Vaughan, A. H. Price, and M. Davies. 1969. Dielectric Properties and Molecular Behavior. Van Nostrand Reinhold Company Ltd., London. 16-23. Lakowicz, J. R. 1983. Principles of Fluorescence Spectroscopy. Plenum Press, New York. 128. Lami, H., and N. Glasser. 1985. Indole's solvatochromism revisited. J. Chem. Phys. 84:597-604. Levy, R., J. D. Westbrook, D. Kitchen., and K. Krogh-Jespersen. 1991. Solvent effects on the adiabatic free energy difference between the ground and excited states of methylindole in water. J. Phys. Chem. 95:6756-6758. Lippert, E. 1957. Z. Electrochem. 61:962. Liptay, W. 1976. Ber. Bunsen-Ges. Phys. Chem. 80:207-217. Mataga, N., Y. Torihashi, and K. Ezumi. 1964. Electronic structures of carbazole and indole and the solvent effects on the electronic spectra. Theoret. Chim. Acta. 2:158-167. Mathies, R. A. 1974. Experimental and Theoretical Studies on the Excited Electronic States of Some Aromatic Hydrocarbons through Electric Field Perturbation and through Chemical Substituents (Ph.D. thesis, Cornell University). University Microfilms International, Ann Arbor, MI. 99-139. Meixner, A. J., A. Renn, and U. P. Wild. 1992. Spectral hole-burning and Stark effect: a centrosymmetric molecule in polymers of different dielectric constants. Chem. Phys. Lett. 190:75-82. Muino, P. L., D. Harris, J. Berryhill, B. Hudson, and P. R. Callis. 1992. Simulation of solvent dynamics effects on the fluorescence of 3-methylindole in water. J. R. Lakowicz, editor. Proc. SPIE. 1640: 240-251. Oh, D. H., M. Sano, and S. G. Boxer. 1991. Electroabsorption (Stark effect) spectroscopy of mono- and biruthenium charge-transfer complexes: measurements of changes in dipole moments and other electrooptic properties. J. Am. Chem. Soc. 113:6880-6890. Press, W. H., W. T. Vetterling, S. A. Teukolsky, and B. P. Flannery. 1992. Numerical Recipes in Fortran, 2nd ed. Cambridge University Press, New York. Rehms, A., and P. R. Callis. 1987. Resolution of La and L. bands in methyl indoles by two-photon spectroscopy. Chem. Phys. Lett. 140:83-89. Valeur, B., and G. Weber. 1977. Resolution of the fluorescence excitation spectrum of indole into the 'La and 'L, excitation bands. Photochem. Photbiol. 25:441 444. Vauthey, E., K. Holliday, C. Wei, A. Renn, and U. P. Wild. 1993. Stark effect and spectral hole-burning: solvation of organic dyes in polymers. Chem. Phys. 171:253-263. Walker, M. S., T. W. Bednar, and R. Lumry. 1967. Exciplex studies. II. Indole and indole derivatives. J. Chem. Phys. 47:1020-1028. Westbrook, J. D., R. M. Levy, and K. Krogh-Jespersen. 1992. Simulation of photophysical processes of indoles in solution. J. R. Lakowicz, editor. Proc. SPIE. 1640:10-19.
Biophysical Journal Volume 68 April 1995 1583-1591
Stark Effect Spectroscopy of Tryptophan Daniel W. Pierce and Steven G. Boxer Department of Chemistry, Stanford University, Stanford, California 94305 USA
ABSTRACT The change in permanent dipole moment (IA1p I) for the transition from the 1La state to the ground state of tryptophan is the key photophysical parameter for the interpretation of tryptophan fluorescence spectra in terms of static and dynamic dielectric properties of the surrounding medium. We report measurement of this parameter by means of electric field effect (Stark) spectroscopy for N-acetyl-L-tryptophanamide (NATA) in two solvents, the single tryptophan containing peptide melittin, and 5-methoxytryptophan. The values ranged from 5.9 to 6.2 ± 0.4 Debye/f for NATA and melittin, where f represents the local field correction. The 1Lb AlP was much smaller. Application of Stark spectroscopy to these chromophores required decomposition of the near-UV absorption into the 1La and 'Lb bands by measurement of the fluorescence excitation anisotropy spectrum and represents an extension of the method to systems where band overlap would normally preclude quantitative analysis of the Stark spectrum. The results obtained for 5-methoxytryptophan point out limitations of this method of spectral decomposition. The relevance of these results to the interpretation of steady-state and time-resolved spectroscopy of tryptophan is discussed.
INTRODUCTION Measurement of steady-state and time-resolved fluorescence of tryptophan is widely used for characterization of peptides and proteins, and results are routinely interpreted in terms of the static polarity and heterogeneity of the local environment as well as the dynamic properties of the environment and the emitting residue (Beecham and Brand, 1985). The key photophysical parameter for the interpretation of fluorescence in terms of local polarity is the change in permanent dipole moment ( A'i ) on absorption. The classical method for measuring this quantity is to measure the fluorescence spectrum in a series of solvents of known polarity. The difference dipole is given by the Ooshika-Lippert-Mataga relation
(Lippert, 1957):
(AV)- 2("ie pg)
(r
I
n
2
1
(1)
where (A v) (A vA) - (AVF), and (AvA) and (AvF) are the shifts of the 0-0 absorption and fluorescence transition energies, respectively, due to electrostatic interactions with the medium. These shifts are measured relative to the values obtained in a nonpolar solvent. The quantities Me and Mg are the excited- and ground-state permanent dipole moments, and A'i = Me- Mg. E0 is the static dielectric constant, n is the refractive index, a is the cavity radius of the solvent cavity (typically assumed to be spherical) required to accommodate the chromophore, and h is Planck's constant. This equation remains extremely useful but suffers from the same shortcomings as all relations based on a continuum dielectric model of a solvent: all perturbations due to specific inter-
Received for publication 28 July 1994 and in final form J0 January 1995. Address reprint requests to Dr. Steven G. Boxer, Department of Chemistry, Stanford University, Stanford, CA 94305-5080. Tel.: 415-723-4482; Fax: 415-723-4817; E-mail: [email protected]. C) 1995 by the Biophysical Society
0006-3495/95/04/1583/09 $2.00
actions between solute and solvent (i.e., ground- or excitedstate complex formation) and to the finite size of the solvent molecules are ignored. Such effects will adversely impact the accuracy of the determination of AM I . In addition, Me is assumed to be the same in the Franck-Condon excited state and the relaxed emitting state, and obtaining reliable absorption and fluorescence spectra of dipolar or charged molecules in a truly nonpolar reference solvent can be challenging because of limited solubility and aggregation. A better method, in principle, is to measure the effect of an externally applied electric field on the electronic transition (the Stark effect; Liptay, 1976). The electric field changes the wavelengthdependent molar absorption coefficient in a manner dependent on A' I, the change in the average polarizability (TrAa), and the internal angle ; between the transition moment 'm and AM,. Here we report the results of Stark effect and fluorescence excitation anisotropy studies at low temperature (77-85 K) on N-acetyl-L-tryptophanamide (NATA) in ethanol (EtOH) or 50% v/v glycerol/water glass, melittin in 50% glycerol/water, and 5-methoxytryptophan in EtOH. These molecules were selected to examine the sensitivity of the spectra to different environments and to chemical substitution. The near-UV absorption of tryptophan is believed to consist of two nearly totally overlapping T-u7T* transitions, labeled 'La and 1Lb in the Platt notation (Valeur and Weber, 1977). The formalism for the analysis of Stark spectra developed by Liptay (1976) rests on the assumption that the absorption lineshape is that of a single, isolated electronic transition with wavelength-independent electrooptic properties. This assumption is necessary because the Stark spectrum is expressed in terms of sums of derivatives of the absorption lineshape. However, there is no fundamental obstacle to analysis of the Stark spectrum of overlapping transitions as long as there is some means of decomposing the total absorption into its constituent parts. In the case of
1584
Biophysical Journal
tryptophan, the observation that the fluorescence originates solely from 'La and the near orthogonality of the 'La and 'L4 transition moment directions provides a means for accomplishing such decomposition, performed originally by Valeur and Weber (1977). The emission anisotropy at differing excitation wavelengths is measured and interpreted in terms of the fractional absorption accounted for by each transition moment at each excitation wavelength. 5-Methoxytryptophan was selected because substitution of electronwithdrawing moieties at position 5 is thought to selectively lower the energy of the '4 transition and switch the order of 0-0 energies in both absorption and fluorescence without perturbation of 'La, and previous work has indicated that this is the only methoxy derivative that retains orthogonality of the transition moments (Eftink et al., 1990).
MATERIALS AND METHODS NATA was a gift of David J. Lockhart. Melittin and 5-methoxytryptophan were purchased from Sigma Chemical Co. (St. Louis, MO) and used without further purification. The glycerol was Mallinkrodt AR grade and the EtOH was Aldrich spectrophotometric grade.
Stark and absorption spectra The Stark apparatus used in these experiments has been described elsewhere (Boxer, 1993). Light from a 450-W Xe arc lamp (Spex) was dispersed through a Spex Minimate ¼/4-m monochromator (0.9-nm resolution), collimated, passed through an air-spaced Glan-Taylor polarizer and a strain-free quartz optical dewar containing the sample immersed in liquid nitrogen, and detected by a photomultiplier tube on which the resistor network had been altered so as to bypass the last few dynodes and decrease the noise. The photomultiplier signal was converted to a voltage and amplified using a home-built converter/amplifier and fed into a lock-in amplifier (Stanford Research Systems SR530). Samples consisted of 15 ,ul of solution between two quartz slides held apart by a 25-,um Kapton spacer. The quartz slides were rendered conductive by means of an evaporated layer of nickel -7.5 nm thick, corresponding to a transmittance of about 50% per slide in the near UV and visible and a resistance of not more than 1000 Q. 400-Hz sinusoidal AC electric fields with peak-to-peak voltages from 2 to 4.5 kV were produced by a custom-built power supply and applied to the sample. The raw data are the intensity of the signal at the photomultiplier, I(v), and the amplitude of its modulation at twice the field frequency, AI(v). The Stark spectrum is AI(v)/(I(v)-k), where k is a constant baseline voltage from the amplifier. Because the electric field modulation is small (1150 K. Although this energy will rapidly flow into a large number of low-frequency vibrational modes, perhaps enough local heating occurs so that small rotations of the molecule may occur. An analogous process could occur on emission, but by the time a vibrationally hot ground state is created the photon is gone and any rotation of the ground-state molecule will not be observed. In either case, the validity of the absorption decomposition given by Eq. 4 must be called into question if such decomposition is carried out over multiple vibronic maxima of ei-
z Fm .
5-methoxytryptophan
ca
0
0.05 V
0
0.0
0.00
0-
32000 33000 34000 35000
-0.1
21000 24000 27000 30000 ENERGY
cm
l
ENERGY cm-1
FIGURE 6 Absorption (A) and fluorescence excitation anisotropy (V) spectra for 5-methoxytryptophan at 77 K in EtOH. The absorption data is identical to that in Fig. 5 except that a smaller range of data is shown. Solid lines do not show fits in this figure.
FIGURE 7 5-Methoxytryptophan singlet (A) and triplet (A) emission and emission anisotropy spectra. The fluorescence and phosphorescence spectra have been independently scaled to a peak intensity of 1. The spectra have been corrected for the wavelength and polarization-dependent efficiency of the detection apparatus.
Pierce and Boxer
1 589
Stark Effect Spectroscopy of Tryptophan
ther transition. For 5-methoxytryptophan, there does not appear to be a clear-cut means of identifying the onset of 'La absorption from r,(v), so we analyze the Stark spectrum of the first vibronic band in the normal manner for an isolated electronic transition. For NATA and melittin, the excitation anisotropy spectra observed are more plausibly attributed to overlapping electronic transitions. To avoid running afoul of assumption 1, we restrict our analysis of the Stark spectra of these molecules to the region containing the first vibronic band of each transition (32,000-35,000 cm-').
DISCUSSION Comparison with solvent-dependence data Mataga and co-workers (1964) estimated Ai for the 'La state of indole to be 5 Debye (D) on the basis of the solvent dependence of the emission spectrum, and Gladchenko and Pikulik (1967) estimated jie to be 5.6 D based on the temperature dependence of absorption and emission spectra. These results were called into question by Walker et al. (1967), who interpreted at least part of the solvent dependence to exciplex formation in protic solvents. Lami and Glasser (1985) undertook a more contemporary and detailed study that resolved several points of confusion in the literature. First, they demonstrated that the identity of the emitting state depends on solvent polarity for indole and several derivatives, which presents a clear problem in the interpretation of solvent-dependent emission data. Second, evidence for ground-state complex formation in all polar solvents tested leading to large shifts in the 'La energy was obtained. Third, solvent effects on absorption and emission were separately evaluated. However, ground-state complex formation precluded independent determination of A'i for 'La absorption and emission, so the assumption of no change in je on relaxation of the Franck-Condon excited state still had to be made. Under this assumption, for the 'La state of 3-methylindole was found to be 5.32 D, and that of the 1Lb state of 5-methoxyindole 2.69 D. The corresponding groundstate values were 2.10 and 2.0 D, leading to A'i values of 3.22 and 1.69 D, respectively. This value for the 'La Api is smaller than that obtained in this work. The difference is too large to be accounted for by any plausible estimate of the local field correction, and must therefore be ascribed to limitations in the accuracy of one or both of the methods being compared.
surements is qualitative at best. Condensed-phase molecular dynamics coupled with similar electronic structure calculations (Muino et al., 1992) indicate an even larger value for the 'La dipole. Chabalowski et al. (1993) performed ab initio calculations on indole including optimization of excited-state geometries and the effects of a reaction field. They found 'La and 1Lb lie values of 6.74 and 3.17, respectively, and a igI of 2.94 D after two iterations at the ground-state geometry and reaction field, although the convergence of the calculation after this number of iterations is unknown. However, the corresponding values for the fluorescence transition at the excited-state geometry and reaction field were 10.4, 4.72, and 5.41 D, respectively. Levy and co-workers have performed extensive vacuum and condensed phase ab initio electronic structure calculations with great attention to the effects of the solvent field on the properties of electronic states (Levy et al., 1991; Westbrook et al., 1992). The change in dipole moment on excitation to the 'La state predicted by their work is 3.1 D. This value represents an ensemble average over many configurations of solvent. The range of values of l for the configurations sampled was 2.4 to 9.0 D (K. Krogh-Jespersen, personal communication), and the excited-state dipole moment was found to increase rapidly to 14 D in simulated water at 300 K as the water reorganized to solvate the excited-state dipole moment, and the excitedstate dipole moment in turn changed in response to the reaction field due to the water. A common theme in all of these computations is the large change in the charge distribution of the 'La state that occurs during the excited-state lifetime in polar solvents. The magnitude of these changes illustrates the displaced nature of the 'La transition and the hazards present in assuming that photophysical properties are independent of time and environment (as is assumed in the classical solvent shift relations.)
Inhomogeneous effects and solvent dependence The Stark lineshape data in this work are analyzed under the assumption that the electrooptic properties of the molecule are not inhomogeneously distributed, or more precisely that the width of the distribution in a parameter is small compared with the magnitude of the parameter. The Stark signal is quadratic in Ali and Aa I, so the quantities determined are actually the square root of the ensemble average of the square of the quantity. For example, if Ali is characterized by a gaussian distribution with width ar and center value POiA the measured value is given by
Comparison with electronic structure calculations Callis (1991) performed extensive semiempirical molecular orbital calculations on indole and was able to reproduce many experimentally observed features. Calculated 'La l values ranged from 3.22 to 7.07 D but were generally about 5 D. A gas-phase value for Api of 5.7 D for 3-methylindole is also reported by Muinlo et al. (1992). However, the relevance of gas-phase calculations to condensed-phase mea-
2 -I AAo1)2/a2 1 '2 r(/ApiI= I-IA01)2/or2 (I -) f oI Ap' (IAAe-(lAA'
f
00
(12)
e
so
(IAp2 (15p1)
_
V&±IA 0-I2 IA"*
(13)
1 590
Biophysical Journal
where angle brackets denote an ensemble average. For instance, if A'I is 5 D and or is 2 D, the measured value would be 5.39 D. If there is a correlation between the value of an electrooptic parameter and the transition energy of a chromophore, the measured value will also depend on the region of the (inhomogeneously broadened) spectrum that is analyzed (Demchenko and Ladokhin, 1988). An interesting aspect of recent condensed-phase electronic structure calculations is the possibility of evaluating such correlations. Large variations in the 'La dipole moment were predicted in several of the studies described above (Muifio et al., 1992; K. Krogh-Jespersen, personal communication), and in one of them (Muifio et el., 1992) a correlation was predicted, with red-absorbing solvent configurations characterized by larger excited-state dipole moments. Because we focus on the red part (although not the extreme low-energy tail that is usually the subject of "red-edge effect" investigations) of the absorption band, this experiment may be sensitive to this bias (Demchenko and Ladokhin, 1988). However, because an entire vibronic maximum is analyzed for each transition, a width of the order of the inhomogeneous linewidth is included in the analysis, and any bias must be correspondingly small. An experimental approach to the correlation of electrooptic parameters with transition energy is provided by the combination of electric field effects with line-narrowing spectroscopy (Meixner et al., 1992; Altmann et al., 1992). The most directly relevant of the available studies is that of Vauthey et al. (1993). For highly polar dyes in highly polar polymer matrices, a substantial wavelength dependence to the amplitude and lineshape of the electric field effect was observed. The lineshape changes indicated that the angle between Ai and 'm changed as a function of wavelength. The data were interpreted as a rotation of Aji brought about by changing relative contributions of "intrinsic" and matrix-induced difference dipole moments to AjI, but we note that rotation ofthe transition moment due to vibronic coupling (see above) would provide an alternative mechanism. The values of Ai recovered for both transitions for NATA in EtOH, NATA in glycerol/water, and mellitin are identical within the errors despite the rather large changes in 0-0 transition energies and inhomogeneous linewidths seen in Figs. 2 and 4. The Stark spectra in Fig. 3 reflect the differences in absorption lineshapes but are otherwise identical. The solventindependence ofthe electrooptic properties is not consistent with significant inhomogeneous distortion of the recovered parameters. If the electrooptic properties are independent of solvent, they should be correspondingly insensitive to different local structures within the same solvent. It is interesting to note that the sample in the least polar solvent, NATA in EtOH, has the most redshifted 'La and 1, absorbance maxima, in contrast to the continuum electrostatic prediction. The 'La linewidths, however, do follow the expected trends with solvent polarity.
Resolution of the 1La and 'Lb transitions The decomposition of the 'La and 'Lb transitions in this work differs from that performed by Valeur and Weber (1977) in
Volume 68 April 1995
that the decomposed spectra were required to fit the Stark spectrum. Since this constraint was present, it was not necessary to assume a value for the limiting anisotropy of either transition, and these were left as free parameters in the fit. In the case of NATA and melittin, the assumption that the relatively flat region in the low-energy tail of the excitation anisotropy spectrum represents nearly pure 'La absorption was born out by the fitting. The limiting anisotropy determined for the 'La transition is less than the theoretical limiting value of 0.4 and indicates that the transition dipole moments on absorption and emission are not completely parallel. The values obtained (0.32-0.35) correspond to angles of 17-21°. Such rotations can derive from vibronic coupling (see above) or nuclear rearrangements during the excited-state lifetime. Since the original experiment of Valeur and Weber (1977), two other experimental approaches to the decomposition of indole absorption spectra have been pursued. Albinsson and Norden (1992) have measured linear dichroism spectra of indole and derivatives in stretched films, and Rehms and Callis (1987) have measured two-photon fluorescence excitation spectra. In addition, resolutions of a variety of derivatives have been performed by the fluorescence excitation anisotropy technique (Eftink et al., 1990). Of these, only the two-photon absorption measurements do not rely directly on wavelength-independent transition moment directions in order to measure relative 'La and 1Lb oscillator strengths. However, two-photon and one-photon lineshapes need not be identical, so there is at present no unambiguous means of decomposing the total absorption over the entire band. Our observations suggest that the available resolutions may not be even qualitatively useful except in the region of the 0-0 bands. As the analysis of Stark spectra depends rather sensitively on the absorption lineshape, it is likely that the largest nonstatistical errors in this experiment are due to violations of the assumptions necessary in the absorption decomposition.
CONCLUSION The most robust conclusion of this work is that A'i for the 'La transition of the indole chromophore in NATA and melittin is about 6.0 D/f. The long-wavelength absorption tail is due entirely to this transition, and the Stark spectrum in this region is dominated by a second-derivative contribution from it. It is therefore impossible to reproduce the experimental data by any other choice of parameters. The other parameters given in Table 1 are accurate within the assumptions necessary to decompose the absorption spectrum, but do not have as unique a relationship to the Stark spectrum. It can be stated that the 'Lb A'i cannot be much larger than the value given. The magnitude of the Stark effect is proportional to the inverse square of the linewidth of a gaussian transition. Given that the linewidth of the 'Lb 0-0 transition is much narrower than that of 'La, a larger A'i for 'Lb would have a
dramatic effect
the
'La
A'
we
on
the total Stark lineshape. The
obtain is
larger
value of than that calculated from
Pierce and Boxer
Stark Effect Spectroscopy of Tryptophan
solvent shifts or predicted by some electronic structure calculations. A value of the local field correctionf > 1.5 would be necessary to produce agreement with these calculated values. This value offis possible, but would be at the upper end of the likely range (Oh et al., 1991). These results affect the interpretation of tryptophan fluorescence properties in several ways. First, if it is accepted that the 'La A'i on absorption is larger than has been appreciated, quantitative conclusions as to the polarity of the environment based on absorption shifts must be rescaled. The situation with regard to fluorescence shifts is less clear, because the dipole moment may change significantly during the excited-state lifetime. Second, the evidence presented here for wavelength-dependent transition moment directions of one or both states must be taken into account in the interpretation of anisotropy-based measurements of conformational mobilities of tryptophan residues. For example, a rapid anisotropy decay component in the total emission could arise from vibrational relaxation, because different vibronic transitions have different transition moment directions. Such a component could easily be misinterpreted as deriving from motions or 'La-1Lb state dynamics. This work was supported in part by a grant from the National Institutes of Health D.W.P. was supported in part by a National Science Foundation predoctoral fellowship and by the Biophysics Training Grant at Stanford
University.
REFERENCES Albinsson, B., and B. Nord6n. 1992. Excited state properties of the indole chromophore: electronic transition moment directions from linear dichroism measurements: effect of methyl and methoxy substituents. J. Phys. Chem. 96:6204-6212. Albinsson, B., M. Kubista, B. Norden, and E. W. Thulstrup. 1989. Near-ultraviolet electronic transitions of the tryptophan chromophore: linear dichroism, fluorescence anisotropy, and magnetic circular dichroism spectra of some indole derivatives. J. Phys. Chem. 93:6646-6654. Albrecht, A. C. 1961. Polarizations and the assignments of transitions: the method of photoselection. J. Mol. Spectro. 6:84-108. Altmann, R. B., I. Renge, I. Kador, and D. Haarer. 1992. Dipole moment differences of nonpolar dyes in polymeric matrices: stark effect and photochemical hole burning. I. J. Chem. Phys. 97(8):5316-5322. Beechem, J. M., and L. Brand. 1985. Time-resolved fluorescence of proteins. Annu. Rev. Biochem. 54:43-71. Bottcher, C. J. F. 1973. Theory of Electric Polarization, 2nd ed, Vol. 1. Elsevier Scientific Publishing Company, Amsterdam. 77. Boxer, S. G. 1993. Photosynthetic reaction center spectroscopy and electron transfer dynamics in applied electric fields. In The Photosynthetic Reaction Center, Vol. II. J. Deisenhofer and J. R. Norris, editors. Academic Press, San Diego. 179-220. Boxer, S. G., A. Kuki, K. A. Wright, B. Katz, and N. H. Xuong. 1982. Oriented properties of the chlorophylls: Electronic absorption spectroscopy of orthorhombic pyrochlolophyllide a-apomyoglobin single crystals. Proc. Natl. Acad. Sci. USA. 79:1121-1125.
1591
Callis, P. R. 1991. Molecular orbital theory of the 'La and '4 states of indole. J. Chem. Phys. 95:4230-4240. Chabalowski, C. F., D. R. Garmer, J. 0. Jensen, and M. Krauss. 1993. Reaction field calculations of the spectral shifts of indole. J. Phys. Chem. 97:4608-4613. Demchenko, A. P., and A. S. Ladokhin. 1988. Red-edge-excitation fluorescence spectroscopy of indole and tryptophan. Eur. Biophys. J. 15: 369-379. Eftink, M. R., L. A. Selvidge, P. R. Callis, and A. A. Rehms. 1990. Photophysics of indole derivatives: experimental resolution of La and Lb transitions and comparison with theory. J. Phys. Chem. 94:3469-3479. Gladchenko, L. F., and L. G. Pikulik. 1967. Determination of dipole moments of indole and tryptophan in excited states. Zh. Prikl. Spektrosk. 6:355-60. Hill, N. E., W. E. Vaughan, A. H. Price, and M. Davies. 1969. Dielectric Properties and Molecular Behavior. Van Nostrand Reinhold Company Ltd., London. 16-23. Lakowicz, J. R. 1983. Principles of Fluorescence Spectroscopy. Plenum Press, New York. 128. Lami, H., and N. Glasser. 1985. Indole's solvatochromism revisited. J. Chem. Phys. 84:597-604. Levy, R., J. D. Westbrook, D. Kitchen., and K. Krogh-Jespersen. 1991. Solvent effects on the adiabatic free energy difference between the ground and excited states of methylindole in water. J. Phys. Chem. 95:6756-6758. Lippert, E. 1957. Z. Electrochem. 61:962. Liptay, W. 1976. Ber. Bunsen-Ges. Phys. Chem. 80:207-217. Mataga, N., Y. Torihashi, and K. Ezumi. 1964. Electronic structures of carbazole and indole and the solvent effects on the electronic spectra. Theoret. Chim. Acta. 2:158-167. Mathies, R. A. 1974. Experimental and Theoretical Studies on the Excited Electronic States of Some Aromatic Hydrocarbons through Electric Field Perturbation and through Chemical Substituents (Ph.D. thesis, Cornell University). University Microfilms International, Ann Arbor, MI. 99-139. Meixner, A. J., A. Renn, and U. P. Wild. 1992. Spectral hole-burning and Stark effect: a centrosymmetric molecule in polymers of different dielectric constants. Chem. Phys. Lett. 190:75-82. Muino, P. L., D. Harris, J. Berryhill, B. Hudson, and P. R. Callis. 1992. Simulation of solvent dynamics effects on the fluorescence of 3-methylindole in water. J. R. Lakowicz, editor. Proc. SPIE. 1640: 240-251. Oh, D. H., M. Sano, and S. G. Boxer. 1991. Electroabsorption (Stark effect) spectroscopy of mono- and biruthenium charge-transfer complexes: measurements of changes in dipole moments and other electrooptic properties. J. Am. Chem. Soc. 113:6880-6890. Press, W. H., W. T. Vetterling, S. A. Teukolsky, and B. P. Flannery. 1992. Numerical Recipes in Fortran, 2nd ed. Cambridge University Press, New York. Rehms, A., and P. R. Callis. 1987. Resolution of La and L. bands in methyl indoles by two-photon spectroscopy. Chem. Phys. Lett. 140:83-89. Valeur, B., and G. Weber. 1977. Resolution of the fluorescence excitation spectrum of indole into the 'La and 'L, excitation bands. Photochem. Photbiol. 25:441 444. Vauthey, E., K. Holliday, C. Wei, A. Renn, and U. P. Wild. 1993. Stark effect and spectral hole-burning: solvation of organic dyes in polymers. Chem. Phys. 171:253-263. Walker, M. S., T. W. Bednar, and R. Lumry. 1967. Exciplex studies. II. Indole and indole derivatives. J. Chem. Phys. 47:1020-1028. Westbrook, J. D., R. M. Levy, and K. Krogh-Jespersen. 1992. Simulation of photophysical processes of indoles in solution. J. R. Lakowicz, editor. Proc. SPIE. 1640:10-19.
